M

lth & edica

ea
Journal of Raghupathi and Raghupathi, J Health Med Informat 2015, 6:2
Journal of H

l In
formatics Health & Medical Informatics DOI: 10.4172/2157-7420.1000188
ISSN: 2157-7420

Research article Open Access

Benchmarking Hospital Performance Using Health Analytics

Viju Raghupathi1* and Wullianallur Raghupathi2
1
Brooklyn College, City University of New York, 2900 Bedford Avenue, Brooklyn, NY 11210, USA
2
Graduate School of Business,Fordham University, New York, NY

Abstract
Background: The objective of this study is to investigate hospital performance using an emerging analytics
approach. Given that hospital care accounts for a large segment of healthcare spending, it is essential that hospital
performance be measured over time to determine whether and where there is room for improvement in some of its
critical success factors, and if there are savings to be found.
Methods: Employing indicators such as hospital cost, in-hospital death rate, length of hospital stay, and the number
of discharges from the hospital, we look at the trends for these indicators over a 10-year period. Data was extracted from
the National Statistics Database of the Healthcare Cost and Utilization Project ([Link]), and Cognos and Tableau
were used as visualization and analysis tools.
Results: Our central finding is that over the 10-year analysis period, U.S. hospitals improved in several areas,
including reduction in length of stay in hospitals, reduction in number of in-hospital deaths, and increase in number of
discharges from hospitals. Despite these improvements, however, the cost of healthcare rose significantly.
Conclusions: We show how healthcare administrators can learn from past performance in determining where to
focus attention and improve outcomes. We also present a global perspective of healthcare and propose how critical it is
for the U.S. to focus on major reduction in healthcare costs, beginning with hospital charges.

Keywords: Hospital performance; Healthcare; Analytics; In-hospital and a decrease in most types of mortality. Nevertheless, compared
death; Length of stay; Discharge; Cost; Analytics; Cognos; Visualization to other OECD high-income countries that have demonstrated a
similar trend, U.S. has not moved up the rankings for critical health
Research Background care indicators such as cost, death occurrence, length of stay, and total
Prior studies make it clear that analyzing hospital performance is number of discharges.
critical to healthcare at national and global levels for lowering costs, A 2013 survey of 11 countries conducted by the Commonwealth
improving patient outcomes, and improving overall quality of care. Fund showed that, compared to other countries, adults in the U.S.
On a global level, the Organization for Economic and Cooperative were more likely to forego healthcare due to the high cost [7]. The
Development (OECD) has determined that 20 to 40% of a country’s United States Health Data for 2013 shows that the spending on health
total health expenditure should be on hospital care. On a national per capita for 2011 was $8508, about two and a half times the OECD
level, the general expectation is that efficiency of hospitals in terms of average of $3339 and twice that of such wealthy European countries
patient treatment and care improves over time [1]. Meanwhile, a study as France and Sweden. The healthcare cost for the U.S is more than
of healthcare across the U.S. shows a range of quality, inviting more that of Norway, the country in the OECD with the second highest-
analysis of internal performance factors to help professionals address income. The total spending on healthcare in the U.S for the years 2000
these quality discrepancies [2]. to 2009 showed a yearly average increase of 4.4%, while the growth
Hospital care accounts for the largest share (32%) of the U.S. rate declined by almost half to 2.1% [4]. Among OECD countries, the
healthcare expenditures, followed by physician/clinical services Commonwealth Fund ranked the U.S. lowest in terms of healthcare
(21%), prescription drugs (10%), nursing home care (6%), home and quality and at the top of the list in terms of cost [5].
healthcare (2%) [3]. Given that this is a large segment of healthcare In this research, we use an analytics framework to analyze changes
spending, it is essential that hospital performance be measured over in hospital performance over time in four key areas: cost of hospital
time to determine whether and where a) there is room for improvement stay, length of hospital stay, the number of hospital discharges, and the
in some of its critical success factors, and b) there are savings to be number of in-hospital deaths. All individuals are directly or indirectly
found. impacted by health services, making the cost of healthcare a significant
Of industrialized nations worldwide, the U.S spends the most on societal factor. We identified several key indicators for hospital
healthcare and the most on healthcare per capita, even though the performance, including reduction in hospital length of stays, reduction
quality of healthcare by specific measurements remains relatively low.
In 2010, the U.S. spent one and half times as much as any other country *Corresponding author: Viju Raghupathi, Brooklyn College, City University of
on healthcare and twice the average of the Organization for Economic New York, 2900 Bedford Avenue, Brooklyn, NY 11210, USA, Tel: 732-485-7952;
and Cooperative Development [4]. Continuing this trend, in 2011, E-mail: VRaghupathi@[Link]
aggregate U.S. hospital spending was $387.3 billion, a 63% increase Received March 18, 2015; Accepted April 22, 2015; Published April 27, 2015
from 1997; and the average cost of hospital stay per person was $10,000,
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance
a 47% increase from 1997 [5]. The overall healthcare spending was $2.8 Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-7420.1000188
trillion, a 3.7% increase from the previous year and constituting 17.9%
Copyright: © 2015 Raghupathi V. This is an open-access article distributed under
of the gross domestic product [6].
the terms of the Creative Commons Attribution License, which permits unrestricted
Over the past years, the U.S has seen an increase in life expectancy use, distribution, and reproduction in any medium, provided the original author and
source are credited.

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 2 of 13

in in-hospital deaths, reduction in or prevention of an increase in important of these is the performance of its healthcare system, not the
the cost of healthcare, and optimization of the rates of admittance in costs associated with it.
emergency departments/other hospitals/long term care. For each of
these key performance indicators, benchmarks were developed for Methods
the sake of comparison. The length of stays and in-hospital deaths are Our research involves the use of analytics in the healthcare domain
compared for conditions or procedures that require similar stay or have and adopts a framework of health analytics that is based on the general
similar deaths rates. The cost of healthcare is compared across various framework of business intelligence and data warehousing [10]. The
conditions or procedures. The growth trend in cost is rationalized and framework is generic enough to be applied to any healthcare context
compared to the overall growth in cost and change in macroeconomic in which analytics is deployed. It includes the components of data
conditions. Admittance rates for various conditions and/or procedures collection, data transformation, analytics platform and tool selection,
are compared, and optimization rates established. We looked for and analytics applications. Figure 1 depicts our analytics framework.
combinations of indicators that generate meaningful results and offer
insight on hospital performance. The expectation is that hospitals will Typically, healthcare data is aggregated from several sources, such as
become more efficient, providing appropriate patient services without hospitals, clinical laboratories, radiology centers, insurance companies,
a dramatic increase in costs over time. and public health systems (e.g. CDC, HHS, WHO). Due to the disparate
nature of the sources, the data may lack uniformity in representation
In the domain of healthcare, analytics is an excellent decision and coding. The raw data, therefore, needs to be reconciled in terms
support technology. It enables physicians, nurses, health officials, health of structure and format, to be readied for analysis. The process of ETL
policy makers, and other healthcare entities to make better and faster (extraction-transformation-loading) readies the data for analytics.
health decisions [8]. We show in a trend analysis how the indicators of Once the data are prepared, they are loaded into a data warehouse
hospital performance function. The data are extensively analyzed using managed by one or more warehouse servers, known as relational
the business intelligence and visualization tools of Tableau and Cognos database management systems (RDBMS). Using RDBMS, sophisticated
Insight. Relationships between indicators are established using the queries that use query structures, optimizations, and query processing
advanced statistical modeling technique of linear regression. Through techniques are performed on the data in the data warehouse. Large data
our analysis, we offer empirical and conceptual insights to hospitals in warehouses sometimes have multiple RDBMSs that run in parallel,
maximizing their operational efficiency. performing multidimensional querying and analyses at high speeds,
Hospital performance a process referred to as online analytical processing (OLAP). Some
analytics operations in OLAP include filtering, aggregation, drilling
According to the national study on in-patient hospital stays by the down and pivoting. In addition to querying, reporting servers facilitate
Agency for Healthcare Research and Quality, there were 38.6 million definition, execution and the generation of accurate reports used for
hospital stays in 2011, an 11% increase from 1997. During this time decision-making. Supplementing the reporting and querying servers
frame, total hospital costs rose to $387.3 billion, a 63% increase. The are the data mining engines that further enable multi-dimensional
mean charges per stay – the amount a patient pays to the hospital for OLAP (MOLAP), allowing users to define and publish cubes (of data)
all charges relating to the stay, including room, nursing and tests – was for high-speed and sophisticated processing.
$35,400 in 2011, more than twice the amount in 1997. Meanwhile, the
number of discharges against medical advice increased 41% over the Data collection
years from 1997 to 2011 [5]. For our study, hospital performance data was downloaded from the
Since the passage of the Patient Protection and Affordable Care National Statistics Data of the Healthcare Cost and Utilization Project
Act in 2010, more attention has been given to efficient delivery of website [Link]. The HCUP encompasses the largest collection
healthcare by healthcare administrators, continuous improvement in of longitudinal hospital care data in the U.S. ([Link]
quality of care and patient safety, promotion of health information [Link]/). As shown in Table 1, the hospital performance indicators
technology (electronic medical records), and reduction of per capita included in our analysis were in-hospital deaths, hospital cost, lengths
spending on healthcare. A report by Health Grades Inc. on hospital of stay in hospital, and patient discharges – all crucial in determining
performance in over 4500 U.S. hospitals (relating to 31 of the most the performance of hospitals in the overall healthcare system. The unit
common inpatient procedures and conditions during the years 2010 of data collection for the indicators shown in the table was their use per
through 2012) shows that there are quality disparities within hospitals 100,000. The principal CCS principal diagnosis category, CCS principal
among different procedures/conditions, as well as between hospitals diagnosis category name, and year were used as filters.
within local services [9]. The report also suggests the impact that
patient mortality (in-hospital deaths) and patient complications have
on healthcare cost, and it advances the theory that the root causes of
cost in terms of mortality and complications can be reduced through
the use of minimally invasive surgical techniques [9].
At a global level, healthcare expenditures show considerable
variation across countries in terms of per capital spending and other
trends. (Note that except for Mexico and the U.S., all OECD countries
have universal or quasi-universal health coverage [4]. Between 2009
and 2011, many countries reduced spending in an effort to cut the
budget deficit or government debt. Canada and the U.S., however,
increased healthcare spending during this period. That said, the overall Source: Raghupathi and Raghupathi, 2013a
health of a population is influenced by many factors, but the most Figure 1: Analytics framework.

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 3 of 13

Indicators Definition

CCS principal diagnosis category Clinical Classifications Software principal diagnosis category

CCS principal diagnosis name Clinical Classifications Software principal diagnosis name

Year Year the data relates to

Total number of discharges Number of patients leaving the hospital after receiving care

Rate of discharges per 100,000 persons Number of patients leaving the hospital after receiving care per 100,000 persons
Number of nights the patient remained in the hospital for his or her stay. A patient admitted and discharged
LOS (length of stay)
on the same day has a length of stay equal to 0.
Amount of hospital billing for the entire hospital stay (not including physician fees). We use the terms costs
Charges, $ (mean)
and hospital charges interchangeably.
Aggregate charges, $ (the "national bill") Sum of all costs for all hospital stays.
Admitted from emergency department Number, Rate Number and rate of patients admitted from the emergency department

Admitted from other hospital Number, Rate Number and rate of patients admitted from other hospitals

Admitted from long term care number, Rate Number and rate of patients admitted from long-term care

In-hospital deaths Number/Rate Number and rate of patient deaths during hospital stay
Routine discharge Number/Rate Number and rate of patients routinely discharged
Discharge to another short-term hospital Number of Patients discharged to another short-term hospital
Discharge to another institution Number of Patients discharged to another healthcare institution

Discharge to home health care Number of Patients discharged to home health care
Against medical advice (AMA) number Number of patients who leave the hospital against the advice of their doctor.
Table 1: Hospital performance indicators.

Using analytics, we identified trends in the significant indicators of Analytics applications

hospital performance over the time period 1997 to 2006. Among other
questions, we asked: Using the selected tools, different types of analytics was performed
in the context of queries, reports, online analytical processing and
- Have hospital costs changed over the 10-year time period data mining. The data that are cleansed and readied are then loaded
from 1997 to 2006? into Cognos and Tableau for analysis. We utilized Cognos Studio
tools for querying, reporting. We utilized the data mining techniques
- Has there been a reduction in hospital deaths over the 10-
of ranking, association and visualization. We also used the online
year time period 1997 to 2006?
analytical processing functions of filtering, aggregation, drilling down,
- Has there been a reduction in the length of stay in the hospital and pivoting of data. For display, we used the dual display feature of
over the 10-year time period 1997 to 2006? chart and table. We also relied heavily on Tableau’s visualization using
scatter plots, highlight tables, and trend line features. The results of our
- Has there been a reduction in the total number of discharges
analysis offer insight into the trends in hospital performance and help
over the 10-year time period 1997 to 2006?
us better assess the state of healthcare performance at a national level.
Data transformation
Results and Discussion
In order to analyze raw data extracted from the Agency for
Healthcare Research and Quality it must be reconciled in terms of Various analytical techniques were applied to reveal associations
structure and format. Quality issues were corrected by integrating, between indicators and to uncover patterns that offer insights into our
cleansing, and standardizing the data through the steps of extract, propositions. Data was explored and analyzed in several ways, and we
transform, and load [11]. generated a total of 13 charts and tables. What follows is a discussion of
our results for the different analyses.
Analytics platform and tool selection
Healthcare charges and cost
Once the data are transformed, a suite of tools is utilized to perform
different types of analytics. It is typical to have a combination of business The trends in the mean healthcare costs for the U.S. displayed in
intelligence tools for analytics. We selected Cognos Studio and Tableau Figure 2 show that healthcare costs, on average, have doubled in the
as the business intelligence tools. The Cognos Studio platform is very years between 1997 and 2006, from $413,482 to $908,172. There is
effective for time series analysis of data. Tableau, especially practical a general increase in cost across all diseases, but the largest increase
for its trend line function, was used for data visualization. Analysis appears to be for respiratory distress syndrome, followed by spinal cord
consisted of manipulating different indicators to find significant trends injury ($790,573) and low birth weight and fetal growth retardation
and patterns in the data. Specifically, we deployed linear regression for ($726,836). The diseases with the least cost are cancers of gastro
the Tableau trend line application. intestinal organs and peritoneum.
To optimize performance and minimize cost given the wide range

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 4 of 13

mentioned above, hospitals would do well to monitor the utilization raising patient costs. The high charges may indicate that hospitals
of their resources in terms of the cost for the various diseases. If are either incurring higher costs or increasing their profit margin per
the portfolio of every hospital is developed accurately, it may reveal patient (or both).
opportunities within the healthcare system to offer specializations.
From a hospital’s perspective, a focus on one or more specialized areas Hospital charges and against medical advice
provides an opportunity to increase prices for those specialized services Against medical advice (sometimes known as discharge against
and see decreasing costs associated with error. medical advice) is a healthcare term that refers to a patient leaving the
hospital against the advice of his or her doctor. Even though leaving a
Healthcare charges and number of discharges
hospital prematurely may be counter-productive to the patient’s well-
An interesting finding is that the total healthcare cost in dollars being, a patient (or their authorized representative) has the right to
is increasing at a faster rate than the number of discharges (Figure 3). decline recommended treatment and to leave. Healthcare data suggests
that patients discharged against medical advice have a four times higher
Since 2004, while the total number of discharges remained
rate of readmission within 30 days than other patients [12].
relatively stable, total charges have increased by 21%, as shown in
Figure 3. In fact, from 1997 to 2006, the charges have doubled. This We found a statistically significant correlation between aggregate
finding indicates that hospitals are becoming more expensive without charges and against medical advice number for congestive heart
offering any incremental discharge benefit to patients. failure. Figure 5 shows that the number of discharges against medical
advice for congestive heart failure almost doubled from 6710 to 11,433
Hospital charges and in-hospital death rates between the years 1997 to 2006. The aggregate charges also more than
During the analysis time frame of 1997 to 2006, in-hospital death doubled between these years, from $11,819,073,303 to $32,762,953,824.
rates decreased by 1%, while the hospital charges doubled (increase of The positive correlation between the two indicators suggests that if
101%), illustrated in Figure 4. hospitals minimize the increase in charges, the rate of discharges may
also decrease. In the case of congestive heart failure, hospitals should
Meanwhile, as Figure 4 also shows, in-hospital death rate remained try and lower the charges so that more patients can avail themselves of
at 8% through 2003 and decreased to 7% from 2004 to 2006. Again, treatment recommended by their doctors. It is no great cause and effect
since 1997, death rates have decreased and overall costs have increased. leap of understanding from the fact of increased cost of a treatment
This trend indicates that hospitals have had fewer deaths meanwhile

Figure 2: Trends in healthcare costs.

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 5 of 13

Figure 3: Aggregate charges vs. number of discharges.

Figure 4: Average charges vs. in-hospital death rates.

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 6 of 13

Figure 5: Average charges vs. against medical advice.

to the implication that patients may refuse treatment (and leave the disease stage, and severity of illness during admission. Transparency in
hospital) against doctor’s advice due to inability to meet those high hospital pricing is an alternative to keeping prices in check.
charges.
Death rate and diseases
In-hospital death rate and length of stay
As shown in Figure 8, there does not appear to be a correlation
Figure 6 shows the analysis for in-hospital death rate and the length between death rates and the number of diseases.
of stay in the hospital.
For instance, the three most expensive diseases do not show
We found that there is a statistically significant positive correlation the highest death rates. Pneumonia and acute cerebrovascular, two
between the length of stay and in-hospital death rate for liver cancer: expensive diseases, have low death rates. Cardiac arrest, the least
the longer the stay, the higher the death rate. From 1997 to 2006, expensive disease, has one of the highest death rates.
the death rate for liver cancer decreased by 15% from 0.178 to 0.15.
However, during this time frame, the length of stay also decreased by Rate of discharge and cost
7% from 7.379 to 6.851. Thus, even though the death rate decreased Figure 9 shows the rate of discharge for various diseases.
over these years, it’s important to recognize that this decrease can be
attributed partly to the shortening in the length of stay at the hospital, Live births have the highest rate of discharges of all, at about 1,416
which naturally lowers the chances for in-hospital deaths. As far as per 100,000, and the rate increased 4% from 1997 to 2006. If this rate
mortality goes, hospitals should aim to reduce the in-hospital death continues, live births will have an impact on the overall discharge rate
rate by improving the quality of service for diseases with a history of for all diseases (and reasons for hospital stays). Coronary atherosclerosis
high mortality. From our analysis, because liver cancer has a high in- has the second largest number of discharges at 448 per 100,000. Fluid
hospital death rate, a better strategy for hospitals would be to focus on and electrolyte has the lowest rate of discharges at about 187 per
improving the quality of treatment for cancer patients and lower the 100,000. Controlling the discharge rate is important to addressing
in-hospital death rate. Obviously, cancer carries with it an inherent risk healthcare expenses. Premature discharges may trigger readmissions,
of in-hospital mortality. Hospitals should periodically evaluate patient adding to cost. Hospitals should monitor closely the discharge rate in
portfolios and, when appropriate, transfer patients to other hospices or an effort to control long-term costs and improve the quality of service.
maintenance facilities for continued care. In-hospital death number and diseases
Hospital charges across diseases Figure 10 shows the distribution of in-hospital death numbers for
As shown in Figure 7, hospital charges vary across diseases. The various diseases.
mean charges are about $150,000 while the range is from almost $0 to As shown in Figure 10, septicemia and pneumonia are conditions
$900,000. Many factors may cause price disparities across diseases, or that see a high number of deaths or a high risk for death while in the
even across regions. These include hospital location, patient’s health,

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 7 of 13

Figure 6: In-hospital death rate and length of stay (LOS).

Figure 7: Average charges across diseases.

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 8 of 13

Figure 8: Death rate and number of diseases.

Figure 9: Rate of discharge.

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 9 of 13

Figure 10: In-hospital death number and diseases.

hospital. The chart indicates that septicemia increased dramatically discharges to short-term hospitals, rehabilitation centers and nursing
from about 66,000 in 2003 to about 110,000 in 2006. Gastrointestinal homes (and other institutions), and home health care in the year 2006.
hemorrhage has the lowest in-hospital death number, about 126,000,
That year, osteoarthritis had the highest number of patients
and also shows a decreasing trend from 1997 to 2006. Fluid and
discharged to home care and to rehabilitation/nursing homes. Where
electrolyte disorder is another condition that shows a decreasing trend
short-term hospitals are concerned, coronary atherosclerosis had the
during this timeframe. To reduce in-hospital death rates and develop
highest number, and gastrointestinal hemorrhage had the lowest; for
solutions for improving the quality of patient care during stays,
home care, osteoarthritis had the largest number of discharges, and
hospitals need to consider the variation in these death number trends
fluid and electrolyte disorder had the lowest; and for discharge to
and determine which diseases to focus on.
nursing home or rehabilitation centers, osteoarthritis had the largest
Length of Stay, admitted from long-term and admitted from number while diabetes mellitus with complications had the lowest.
emergency This information is useful in discharge planning evaluation, which
determines a patient’s care needs after leaving the hospital setting.
Figure 11 shows the trend for Length of Stay, the number admitted Hospitals should be prepared to transfer medical records with ease to
from long-term care and the number admitted from emergency care. other institutions, as well as perform appropriate needs assessments,
The analysis of the model for Length of Stay and the number depending on where the patient is being transferred.
from the two modes of admission - long-term care and emergency Length of stay and against medical advice
care - was significant (R2 is close to 1; p-value<0.0001) (Figure 11). In
Figure 11, the numbers admitted from long-term care are in columns, Figure 13 shows the distribution of the length of stay and against
and the numbers admitted from emergency care are in rows. CCS medical advice.
principal diagnosis is the filter. For acute cerebro-vascular condition, The model for length of stay and the number discharged against
the number admitted from long-term care and from emergency care medical advice has been significant for the years 1997 to 2006 (R2=0.6;
showed a significant relationship. For others, such as abdominal hernia, p<0.05). As shown in Figure 13, the total numbers for Length of Stay
abdominal pain, acute and unspecified renal failure, acute bronchitis, and those leaving Against Medical Advice have decreased over this 10-
acute myocardial infraction, and affective disorders, the length of stay year period, demonstrating considerable improvement in quality of
and the number admitted from long-term care and emergency care hospital service.
show a significant relationship. In general, the number admitted from
long term care and emergency care decreased considerably with the Total number of discharges and in-hospital deaths for each
same length of stay for these diseases. This can be attributed to the fact disease
that hospitals have been able to efficiently control and/or treat these
particular diseases. Figure 14 shows the total number of discharges and the number of
in-hospital deaths.
Discharge of patients to different healthcare providers
The model for total number of discharges and in-hospital deaths
Figure 12 shows our analysis of the distribution of patient is significant for the 10-year time period (R2=0.997, p-value<0.0001),

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 10 of 13

Figure 11: Trend line for length of stay, admitted from long-term care and admitted from emergency.

Figure 12: Discharge of patients to other healthcare providers.

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 11 of 13

Figure 13: Length of stay and against medical advice.

Figure 14: Total number of discharges and number of in-hospital deaths.

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 12 of 13

and the relationship between these two numbers, across almost all and other resources for continuous care. The positive correlation
diseases, is also significant. In the case of acute myocardial infarction, between length of stay and the number of discharges against medical
the two indicators showed a decreasing trend over the years. For acute advice indicates that hospitals need to pay attention to determining
and unspecified renal failure, the two indicators showed an increasing the length of stay for the portfolio of diseases for which patients are
trend. Hospitals should monitor these relationships for various diseases hospitalized. Between 1997 and 2011, the number of discharges against
and address how to improve negative trends. medical advice has increased 41% in the U.S., and is dominated by 18-
44 year olds [5]. This number should be addressed and lowered.
Analysis of hospital charges and national healthcare cost
We performed a what-if analysis on the national cost given a Conclusions
hypothetical increase in the hospital charges (Figure 15). Hospitals in the U.S. have improved over the years in terms of
Based on historic data, we found that if the total charges for hospitals reducing length of stay, reducing number of in-hospital deaths, and
increased by 10%, the national cost would increase by over $413 billion. increasing number of discharges from hospitals. However, the cost of
Analysis like this will help hospitals predict future healthcare costs for healthcare has risen significantly during the same period. Escalating
various scenarios and plan accordingly. It can also help identify which healthcare costs impact the national cost, as we have shown in our
specific diseases will have the highest impact on overall costs. analysis.

From our analyses, hospitals can assess the portfolio of charges for Meanwhile, the U.S has been rated very low for the quality of
various diseases and make decisions on optimizing overall costs. Our healthcare service. Hospitals striving to make themselves marketable
results, which show that during the period of analysis, hospitals have to managed care providers and to their patients need to address the
grown more expensive without any incremental discharge benefits negatives, in particular finding ways to improve in the areas of patient
to the patients, have major implications for public and national care and healthcare delivery. Findings from our study indicate that
healthcare, and from them, we can propose specific insights into much is needed to achieve these objectives. Health analytics or clinical
disease management. Hospitals should attempt to reduce the charges analytics will help focus on target areas for improvement and reduce
(for example, for congestive heart failure) to make treatment more the associated direct cost. We show how patients, healthcare providers
affordable to patients. Hospitals might also consider improving the and physicians need to be better informed in terms of what’s working
level of service for diseases that have a history of high death rate (such and what needs improvement.
as liver cancer and septicemea), thereby lowering in-hospital deaths. Scope and limitations
Hospitals can identify factors that positively impact the discharge rate
and also improve the quality of discharge. An ancillary implication Limitations to our study include the fact that it covered only a
is the identification of alternative institutions that patients may be 10-year period. Future longitudinal studies should cover greater time
discharged to and the steps required to ensure a smooth and functional spans so as to reveal more trends and relationships. Second, while we
transition for each patient in terms of the transfer of medical records investigated correlations and trends, we did not include causality across

Figure 15: Aggregate charges vs. national healthcare cost.

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188
Citation: Raghupathi V, Raghupathi W (2015) Benchmarking Hospital Performance Using Health Analytics. J Health Med Informat 6: 188. doi:10.4172/2157-
7420.1000188

Page 13 of 13

the indicators, which future studies can do, thereby helping to identify We also contribute to the methodology of analytics in healthcare.
and target factors that cause increase in healthcare costs and/or decrease Analytics offers a perfect method for mining and analyzing the
in healthcare performance. Third, we utilized a set of indicators that immense amount of data available in the healthcare sector. Our study
were available in the database of the Agency for Healthcare Research compliments the literature of empirical work that deploys an analytic
and Quality. There may be other indicators that better explain or offer ap­proach.
more valuable insight into the phenomenon of hospital performance.
Finally, our study helps patients, the most important aspect of
Lastly, we did not account for or remove outliers from our data set,
the healthcare equation. Patients recognize that hospital performance
and there were some significant outliers relative to other data. In future
varies across procedures. They can estimate a specific hospital’s
studies, the outliers could be removed to understand whether results
performance based on the charges, mortality rates, complications, and
would differ greatly.
other indicators. This is critical since healthcare costs are ultimately
Contributions and policy implications passed on to the patients in the form of higher insurance premiums,
deductibles, and co-payments. When patients are empowered by
Despite limitations, our study contributes to healthcare in several information, they can make appropriate healthcare choices as to
ways. The analysis provides an understanding of how hospitals whether they opt to receive treatment at a particular facility or choose
performed over a recent 10-year period and whether progress is being not to avail themselves of any treatment because of the high cost.
made in patient care. Trends over time for key performance indicators
References
– such as total number of discharges, length of stay, dollar hospital
charges and in-hospital deaths, – help assess whether progress has 1. Hollingsworth B (2003) Non-parametric and parametric applications measuring
been made to improve treatment and/or death rate for certain diseases. efficiency in health care. Heathc Manage Sci 6: 203-218.
Certainly the healthcare system in the U.S. has progressed over the 2. McGlynn EA, Asch SM, Adams J, Keesey J, Hicks J, et al (2003) The quality of
years, but it also has become more expensive. In addition, quality of health care delivered to adults in the United States. New England J Med 348:
healthcare requires attention. 2635-2645.

3. Gutierez C, Saloner R, Ranji U (2007) U.S. Healthcare costs: Background

Hospital management and health professionals can use this
Brief. Medicare Patient Management
analysis to understand how certain measures have changed in 10 years
and focus on improving the weak areas. In addition, the healthcare 4. OECD Health Data (2013): How does the United States compare. OECD
insurance industry can leverage such analysis for their present and 5. Pfuntner A, Wier LM, Elixhauser A (Nov 2013) Overview of Hospital Stays in
future cost structures. Our results contribute to the arena of healthcare the United States, 2011.
reform, so regulators and government agencies may develop effective 6. World Health Organization (2011): World health statistics 2011. Geneva
policies relevant to hospital management and healthcare delivery.
7. Schoen C, Osborn R, Squires D, Doty MM (2013) Access, Affordability, and
In terms of future research, there are many possibilities. As stated, Insurance Complexity Are Often Worse in the United States Compared to 10
future studies can include a more extensive sample, range of indicators, Other Countries. 32: 2205-2215.
and time span. The indicators from the Agency for Healthcare 8. Kearney AT (2013) A healthy dose of data analyt­ics.
Research and Quality may be used in conjunction with other healthcare
9. American Hospital Quality Outcomes (2014): Healthgrades Report to the
indicators to highlight dimensions that were not captured in this study. Nation: Executive Summary.
In addition, further research can incorporate a global perspective by
comparing hospitals in the U.S. to those in other developed countries 10. Raghupathi V, Raghupathi W (2013): Exploring the relationship between ICTs
and Public Health at a Country level. Int J Healthc Inf Syst Inform 8: 1-22.
and discover similarities and differences among the measures. Such a
perspective will contribute to reducing global disparities and improving 11. Raghupathi W, Raghupathi V (2013): An Overview of Health Analytics. J Health
global health. Med Informs 4:132.

12. Fleeter TB (2012) When a patient leaves against medical advice. Am Acad
Orthop Surg 8.

J Health Med Inform

ISSN: 2157-7420 JHMI, an open access journal Volume 6 • Issue 2 • 1000188